FAT10 and NUB1L bind to the VWA domain of Rpn10 and Rpn1 to enable proteasome-mediated proteolysis

ARTICLE

Received 16 nov 2011 | Accepted 16 Feb 2012 | Published 20 mar 2012 DOI: 10.1038/ncomms1752

FAT10 is the only ubiquitin-like modifier that can target proteins for degradation by the proteasome in a ubiquitin-independent manner. The degradation of FAT10-linked proteins by the proteasome is strongly accelerated by the ubiquitin-like–ubiquitin-associated protein nEDD8 ultimate buster-1 long (nuB1L). Here we show how FAT10 and nuB1L dock with the 26s proteasome to initiate proteolysis. We identify the 26s proteasome subunit hRpn10/s5a as the receptor for FAT10, whereas nuB1L can bind to both Rpn10 and Rpn1/s2. unexpectedly, FAT10 and nuB1L both interact with hRpn10 via the VWA domain. FAT10 degradation in yeast shows that human Rpn10 can functionally reconstitute Rpn10-deficient yeast and that the VWA domain of hRpn10 suffices to enable FAT10 degradation. Depletion of hRpn10 causes an accumulation of FAT10-conjugates also in human cells. In conclusion, we identify the VWA domain of hRpn10 as a receptor for ubiquitin-like proteins within the 26s proteasome and elucidate how FAT10 mediates efficient proteolysis by the proteasome.

1 Division of Immunology, Department of Biology, University of Constance, 78457 Konstanz, Germany. ² Biotechnology Institute Thurgau at the University of Constance, 8280 Kreuzlingen, Switzerland. ³ Division of Biochemistry, Department of Biology, University of Constance, 78457 Konstanz, Germany. Correspondence and requests for materials should be addressed to M.G. or to S.G.K. (email: Marcus.Groettrup@uni-konstanz.de or Stefan.Kreft@uni-konstanz.de).

FAT10 and nuB1L bind to the VWA domain of Rpn10 and Rpn1 to enable proteasome-mediated proteolysis

neha Rani

¹

, Annette Aichem

²

, Gunter schmidtke

¹

, stefan G. Kreft

³

& marcus Groettrup

^1,2

Konstanzer Online-Publikations-System (KOPS) URL: http://nbn-resolving.de/urn:nbn:de:bsz:352-210170

A

central proteolytic pathway in eukaryotic cells is the ubiqui- tin–proteasome system¹. The 26S proteasome is a complex of 2.5 MDa comprising the 19S regulatory particle (RP, also called PA700) and a 20S cylindrical core particle (CP). The RP is required for the binding and unfolding of poly-ubiquitylated pro- teins to enable their degradation. The RP can be subdivided into a base composed of six ATPases (Rpt 1–6) and four non-ATPases Rpn2/S1, Rpn1/S2, Rpn13, and Rpn10 and a lid composed of nine structurally diverse subunits. The major known ubiquitin receptors on the proteasome are the RP subunits Rpn10 (designated hRpn10 or S5a in humans^2,3) and Rpn13 (Adrm1 in humans⁴) while the ATPase subunit Rpt5/S6′ has been cross-linked to poly-ubiquitin chains but evidence for its functional involvement in proteolysis is lacking⁵. Rpn10 is assumed to lie at the interface of the lid and the base⁶. This subunit comprises the N-terminal von Willebrand A (VWA) domain⁷ and C-terminal ubiquitin interacting motifs (UIM1 and UIM2, together referred as UIMs)^8–10. Polyubiquitin chains which are covalently attached to the lysine of substrates inter- act with Rpn10 via the ubiquitin interacting motifs (UIMs)^11–13 and with Rpn13 via the pleckstrin-like receptor for ubiquitin (Pru) domain^4,14. In addition, there are further receptors for ubiquitin conjugates belonging to the ubiquitin-like–ubiquitin-associated (UBL–UBA) family of proteins, such as Rad23, Dsk2, and Ddi1, which only temporarily associate with the proteasome¹⁵. This indi- rect pathway involves interaction of ubiquitin chains with their car- boxy-terminal UBA domains whereas their amino-terminal UBL domains interact with the Rpn1, Rpn2 or Rpn10 subunits of the proteasome^16–19.

FAT10 (HLA-F locus adjacent transcript 10) is an 18-kDa pro- tein that has two ubiquitin-like domains with an identity of 29 and 36% to ubiquitin in its N- and C-terminal parts, respectively²⁰. Several ubiquitin-like modifiers have been described but FAT10 is the only one that functions similarly to poly-ubiquitin as a tag for proteasome targeting^21–23. The expression of FAT10 is strongly and synergistically induced in the presence of the proinflamma- tory cytokines interferon gamma (IFN-γ) and tumour necrosis fac- tor alpha (TNF-α)^24,25. Cytokine-treated cells show an extensive high-molecular-weight smear of FAT10 conjugates reminiscent of ubiquitin conjugates, and proteasome inhibition leads to the accu- mulation of FAT10 conjugates^23,26,27. The E1 and E2 enzymes that mediate FAT10 conjugation have recently been identified as UBA6 and USE1, respectively23,26,28–30.

The degradation of FAT10 and FAT10 conjugates is strongly accelerated by the scaffold protein NEDD8 ultimate buster-1 long (NUB1L), which is induced by interferons and contains a N-ter- minal UBL domain and three C-terminal UBA domains^22,31,32. NUB1L binds to FAT10 via its 3 UBA domains and the 26S pro- teasome via hRpn10 (ref. 33). NUB1L may hence serve as a linker between FAT10 and the proteasome. However, FAT10 can also bind directly to the proteasome, and the deletion of the UBA domains of NUB1L interferes with FAT10 binding but not with the acceleration of FAT10 degradation³⁴. The latter effect requires the UBL domain of NUB1L, which is needed for binding to the proteasome, suggesting a ‘facilitator’ function of NUB1L in FAT10-mediated degradation, reminiscent of the function of the Rpn10 VWA domain in Rad23- mediated degradation of poly-ubiquitylated substrates¹⁹.

In this study, we aimed to elucidate the mechanism of accelerated degradation of FAT10 by NUB1L by identifying the interacting sub- units of the 26S proteasome for FAT10 and NUB1L and by charac- terizing their function for FAT10-dependent degradation. We show that FAT10 docked to hRpn10 and NUB1L to hRpn10 or hRpn1.

Both FAT10 and NUB1L bound to the VWA domain of Rpn10 that sufficed to reconstitute FAT10 degradation in yeast. Therefore, the VWA domain of Rpn10 serves as a receptor domain for UBL proteins at the 26S proteasome that mediates FAT10-dependent degradation.

Results

19S RP subunits interacting with FAT10 and NUB1L. To identify the subunit(s) within the 19S RP that mediate FAT10 and NUB1L binding to the proteasome, interaction of each individual 19S RP subunit with FAT10 and NUB1L was analysed in a yeast two-hybrid assay. hRpn10 was the only subunit for which an interaction with FAT10 could be detected (Fig. 1a). hRpn10 also interacted with NUB1L (Fig. 1b), confirming this previously described interaction³³. The previously established interaction of FAT10 with NUB1L was used as a positive control³². Taken together, these results suggest that FAT10 and NUB1L binding to the proteasome occurs via a common 19S RP subunit, hRpn10.

FAT10 binds to hRpn10 and NUB1L to hRpn10 and hRpn1. To test whether the interactions of FAT10 and NUB1L with hRpn10 can be attributed to a direct binding of the proteins, glutathione S-transferase (GST)-pull-down experiments were performed. His- hRpn10 showed specific binding to GST–FAT10 and GST–NUB1L, but not to the GST control (Fig. 2a) confirming a direct interaction of FAT10 and NUB1L with hRpn10. Moreover, we also found spe- cific binding of the yeast Rpn10 protein (scRpn10) to GST–FAT10 and GST–NUB1L (Fig. 2b).

Recently, it was shown that a conserved amino-terminal region of Rpn13, the Pru domain, binds to ubiquitin and also binds to UBL domains of UBL-UBA proteins like hHR23a and hPLIC2 (ref. 4).

Although FAT10 failed to bind to the Pru domain of Rpn13 in an ear- lier study⁴, we re-investigated this possibility with full-length hRpn13

GAL4AD–FAT10 GAL4AD–NUB1L

-TL

-TLH

X-gal 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

17 18 19 20

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

17 18 19 20

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

17 18 19 20 1 2 3 4 5 6 7 8

9 10 11 12 13 14 15 16 17 18 19 20 21

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21

a b

Figure 1 | hRpn10 interacts with FAT10 and NUB1L in a yeast two-hybrid assay. (a) Yeast two-hybrid analysis of the interaction of FAT10 with individual subunits of the human 19s RP. AH109 cells were co-transformed with pGADT7–FAT10 and a pGBD vector encoding one of the human 19s RP subunits fused to GAL4-BD. The colonies represent co-transformants with pGADT7–FAT10 and the indicated pGBD construct: (1) s1/Rpn2, (2) s2/Rpn1, (3) s3/Rpn3, (4) s4/Rpt2, (5) s5a/Rpn10, (6) s5b, (7) s6a/Rpt5, (8) s6b/Rpt3, (9) s7/Rpt1, (10) s8/Rpt6, (11) s9/Rpn6, (12) s10a/Rpn7, (13) s10b/Rpt4, (14) s11/Rpn9, (15) s12/Rpn8, (16) s13/

Rpn11, (17) s14/Rpn12, (18) s15, (19) p55/Rpn5. (20) As positive control, we co-transformed the nmY51 cells with pGADT7–nuB1L and pLexA–

FAT10. A second positive control (21) was the known interaction of nuB1L and hRpn10. Transformants were grown on selection plates containing 3-AT but lacking tryptophan and leucine ( − TL) or tryptophan, leucine and histidine ( − TLH), respectively (upper and middle panels). Confirmation of these results was obtained by X-gal filter assay (lower panel). The experiments have been repeated twice with similar outcomes. (b) Yeast two-hybrid analysis of the interaction of nuB1L with individual subunits of the human 19s RP. The yeast two-hybrid analysis was performed as in (a) except that plasmid pGADT7–nuB1L was used instead of pGADT7–FAT10.

in GST-pull-down assay, and we likewise observed that hRpn13-myc was pulled down neither by GST–FAT10 nor GST–NUB1L.

Taking into consideration that the UBL domain of Rad23 and Dsk2 interact with the leucine-rich-repeat-like domain of the Rpn1 subunit of the 19S RP³⁵, we hypothesized that FAT10 or NUB1L might also interact with hRpn1. Binding assays with recombinant GST–FAT10, GST–NUB1L or GST alone and radiolabelled in vitro translated His-hRpn1 were carried out. Remarkably, His-hRpn1 bound robustly to GST–NUB1L but did not bind to GST–FAT10 or GST alone (Fig. 2c).

To test whether NUB1L binding to hRpn1 depended on the UBL domain or the two UBA domains of NUB1L, GST-pull-down exper- iments with NUB1L-deletion variants were carried out. The dele- tion of the UBL domain (GST–NUB1L∆UBL) resulted in a strong reduction of His-hRpn1 binding (Fig. 2d) indicating a pivotal role of the UBL domain of NUB1L in the binding of His-hRpn1. Moreover, we also investigated whether the UBL domain of NUB1L is essen- tial for the interaction with hRpn10 and, as observed in GST-pull- down assays, the interaction was very weak, if the UBL domain was missing as compared with the interaction when the UBA domains were missing (Fig. 2d). Together, our data show that NUB1L binds to hRpn1 and hRpn10 predominantly via its UBL domain.

The C-terminal UBL domain of FAT10 interacts with hRpn10.

FAT10 consists of two ubiquitin-like domains (Supplementary Fig. S1)²⁰. To determine which FAT10 domain interacts with hRpn10, recombinant His-Rpn10 was incubated with lysates from HEK293T cells transiently transfected with the plasmid carrying HA–FAT10-GFP, HA-N-FAT10-GFP, HA-C-FAT10-GFP, HA-Ub- GFP, and GFP alone. Interestingly, only the C-terminal UBL domain of FAT10 bound to hRpn10 but not the N-terminal UBL of FAT10 (Fig. 2e); full-length HA–FAT10-GFP served as a positive control and HA-Ub-GFP as well as GFP alone did not show any binding to hRpn10 that served as negative controls. Thus, the N-terminal UBL domain of FAT10 binds to NUB1L³⁴, whereas the C-terminal UBL domain binds to hRpn10.

FAT10 and NUB1L interact with the VWA domain of hRpn10. As mentioned above, human Rpn10 contains three domains, the VWA domain and two ubiquitin interacting motifs (UIM1 and UIM2). The UIMs are binding sites for polyubiquitin chains^9,11. To identify the domain within the hRpn10 protein mediating FAT10 and NUB1L binding, GST-pull down assays were performed. In vitro translated full-length His-hRpn10 as well as the His-tagged VWA domain alone (His-VWA) bound to GST–FAT10 and GST–NUB1L, whereas a

GST

+ His-hRpn10

His-hRpn10

His–hRpn10 + + + + +

+ ++ + – – – His-hRpn10 (10% of input) His-scRpn10 (10% of input)

His-hRpn10 His-VW

A His-UIMs His-hRpn10 His-hRpn10 His-hRpn10His-VWA His-VWA

His-UIMs His-UIMs His-VWA His-UIMs UIM2 mut. UIM1 mut. UIM1+2 mut. UIM2 mut. UIM1 mut. UIM2 mut. UIM1 mut.UIM1+2 mut. UIM1+2 mut. UIM2 mut. GST GST–NUB1L GST–FAT10UIM1 mut. UIM1+2 mut.

His-hRpn1 (10% of input)

HA-FAT10-GFP HA-N-FAT10-GFP HA-C-FAT10-GFP HA-Ub-GFP GFP

– – – –– –– – –– – –

+ His–scRpn10

GST–FAT10

His-hRpn10 His-hRpn1

GST–NUB1L

GST GST GST–FAT10 GST–NUB1L

+ GST-

FAT10 +GST-

FAT10 + GST-

NUB1L +GST-

NUB1L +GST + GST

GST–NUB1L∆UBA GST–NUB1L∆UBA

GST–NUB1L∆UBL GST–NUB1L∆UBL

GST–NUB1L GST–NUB1L

GST GST GST–NUB1L∆UBA GST–NUB1L∆UBL

GST–NUB1LGST GST–FAT10 GST–NUB1LGSTGST–FAT10 + hRpn1 GST–NUB1L + hRpn1GST + hRpn1

Commassie staining 10% of inputs

10 % of inputs

10 % of inputs

10 % of inputs

10% of

inputs 10% of

inputs 10% of inputs + His-hRpn1 + His-hRpn10

Coomassie staining

10% of inputs Coomassie staining

10% of inputs

GST–FAT10 GST–NUB1L

b c

f e

d

Pull-down WB: Anti His

IP: Ni beads WB: Anti-GFP

WB: Anti-GFP

55 13095 7255 43 34 26

7255 4334 26 17

95 130 130

95 72 55 43 34

13095 7255

55

His-hRpn10 –+–+–+–+–+

43

43

34

34

26

26 55 43 34 26 17

7255 43 34

95 130 72 55 43 34

26

11 17 11

9572 55 43 34 26 17

9572 55 43 34 26 17

1 2 3 4 5 6 7 8 9 10 11 12

Figure 2 | Direct interaction of FAT10 and NUB1L with the VWA domain of hRpn10 and binding of hRpn1 by NUB1L. (a) Direct interaction of FAT10 and nuB1L with hRpn10. Recombinant GsT, GsT–FAT10 or GsT–nuB1L was incubated with recombinant His-hRpn10 and GsH beads. The interaction was analysed by western blotting with anti-His antibody. 10% of the recombinant proteins are shown. (b) Interaction of scRpn10 with FAT10 and nuB1L.

Recombinant GsT, GsT–FAT10 or GsT–nuB1L was incubated with in vitro transcribed and translated His-hRpn10 or His-scRpn10 along with GsH beads.

The interaction of hRpn10 with GsT–FAT10 served as positive control. (c) Binding of nuB1L to hRpn1. In vitro translated ³⁵s-labelled His-hRpn1 was incubated with recombinant GsT, GsT–FAT10 or GsT–nuB1L on GsH beads and the interaction was monitored by autoradiography; 10% of input of the indicated GsT-tagged proteins is shown on the right. (d) Binding of nuB1L to hRpn1 requires the uBL domain of nuB1L. GsT-pull-down experiments were performed with GsT–nuB1L∆uBL or GsT–nuB1L∆uBA. (e) The C-terminal uBL domain of FAT10 docks to hRpn10. Co-immunoprecipitation of recombinant His-hRpn10 on ni-beads with the lysates from HEK 293T cells transfected with the plasmids shown. Positive interaction was determined by immunoblotting with anti-GFP antibody; 10% of the input of proteins was subjected to anti-GFP western blots on the bottom. (f) FAT10 and nuB1L bind to the VWA domain of hRpn10. GsT-pull-down experiments with GsT–FAT10 or GsT–nuB1L or GsT alone and with in vitro translated radiolabelled full- length hRpn10, the VWA domain, or the two uIms of hRpn10 with wild-type or mutated sequence. Proteins were visualized by autoradiography. 10% of the input of GsT-tagged proteins is shown on Coomassie-stained gels to the right. Experiments were repeated three times with similar results.

no binding was detectable for the two UIMs of hRpn10 (Fig. 2f).

Polyubiquitin chains bind preferably to UIM2 and to a lesser extent to UIM1 (ref. 12). The amino-acid sequence motifs crucial for the binding of polyubiquitin to UIM1 and UIM2 are designated LALAL and IAYAM, respectively¹². To corroborate the above finding, we performed the same experiment with full-length hRpn10 bear- ing a mutated LALAL motif in UIM1 (mutated to AAAAA) and a mutated IAYAM motif in UIM2 (mutated to AAAAA; hereafter referred to as mutated UIMs) and observed that FAT10 and NUB1L can still interact with these mutated proteins (Fig. 2f).

On comparing the amino-acid sequences of FAT10 and ubiqui- tin, we realized that the key residues within ubiquitin required for binding to the hRpn10 UIM domains, Leu8, Ile44 and Val70 nei- ther align with FAT10’s N-terminal nor its C-terminal UBL domain (Supplementary Fig. S1). This observation is in accordance with separate binding sites for FAT10 and ubiquitin within hRpn10. Col- lectively, our results show that FAT10 and NUB1L bind to the VWA domain of hRpn10 in contrast to polyubiquitin chains that bind to the UIM domains of hRpn10.

Rescue of Rpn10∆ yeast with human Rpn10 or the VWA domain.

The role of hRpn10, in particular that of its VWA domain, in the degradation of FAT10 was investigated in Saccharomyces cerevisiae as Rpn10-deficient yeast is viable. We tested whether hRpn10 can functionally substitute for scRpn10 in S. cerevisiae. Yeast cells lack- ing Rpn10 (rpn10∆) show a pronounced sensitivity towards cana- vanine³. Canavanine is an arginine analogue causing misfolding of proteins that are targeted for proteasomal degradation. Consist- ent with an earlier study³, N-terminally FLAG-tagged scRpn10 expressed from a high-copy plasmid lead to a suppression of the canavanine sensitivity of rpn10∆ (NRY5) cells restoring WT-like growth on plates containing 3 µg ml ^{− 1} canavanine, whereas the growth of rpn10∆ cells transformed with an empty plasmid was strongly retarded (Fig. 3). In the same experiment, rpn10∆ cells transformed with the FLAG-hRpn10 expression construct showed growth on a canavanine containing plate comparable to that of rpn10∆ cells reconstituted with scRpn10 (Fig. 3), indicating that hRpn10 can functionally substitute for scRpn10 in S. cerevisiae.

Earlier studies demonstrated that the rpn10-uim mutant or the VWA domain of Rpn10 show wild-type level of sensitivity to cana- vanine^18,36. Whether the VWA domain of hRpn10 alone or hRpn10 with mutated UIMs would be able to suppress the canavanine sen- sitivity of rpn10∆ cells was tested. Interestingly, expression of both, the FLAG-tagged VWA domain as well as the hRpn10 with mutated UIMs, led to a partial suppression of the canavanine sensitivity

(Fig. 3) indicating that the VWA domain alone is able to function- ally reconstitute the Rpn10-deficient yeast proteasome.

FAT10 binds to the proteasome in wild-type but not rpn10∆

yeast. A potential binding of FAT10 with proteasome-associated scRpn10 in yeast was further analysed by glycerol gradient cen- trifugation. Lysates of either WT or rpn10∆ cells expressing N-terminally HA-tagged FAT10 were separated by glycerol gradi- ent centrifugation, and collected fractions were analysed by immu- noblotting (Fig. 4a,b). Free cytosolic Rpn10 was observed in low molecular weight fractions 1–4. Western blots probed with anti- pre6, anti-scRpn10, anti-Rpt1 antibodies showed the presence of 20S and 19S RP in fractions 10–16 and 10–15 in the wild-type and rpn10∆ strains, respectively. We could detect a very low level of RP–CP–RP complex in the fractions from the wild-type yeast (19–20) but not in fractions from rpn10∆ yeast, which is in accord- ance with the previous studies, which showed the dissociation of lid and the base in the rpn10∆ yeast^6,19. The presence of pre6 and Rpt1 in the same fractions showed their existence in the form of a complex (probably RP–CP complex) in the rpn10∆ yeast strain as well. Aminopeptidase I (~600 kDa decamer, AP1) served as a size marker for fractions containing 20S proteasome. In the extract derived from WT cells, HA–FAT10 migrated in the lower and the higher molecular weight fractions (fractions 1–12 and 15–20;

Fig. 4a), whereas it was only found in the lower molecular weight fractions in extracts from cells lacking Rpn10 (fractions 1–8;

Fig. 4b), suggesting the association of FAT10 with proteasome- associated Rpn10. The presence of 20S proteasome in these high molecular weight fractions (fractions 10–15) was further corrob- orated by the presence of hydrolytic activity towards the fluoro- genic substrate Suc-LLVY-AMC (Fig. 4c). The peptidase activity increased in the presence of 0.02% SDS indicating the presence of 20S proteasome, whereas the activity was almost abolished in the presence of the proteasome inhibitor MG132 (Fig. 4c).

The VWA domain of hRpn10 is required for degradation of FAT10. As FAT10 was found to interact with the VWA domain of hRpn10 (Fig. 2f), we investigated whether this interaction would have functional relevance for the degradation of FAT10 in vivo. The degradation rate of ectopically expressed HA–FAT10, in WT and rpn10∆ yeast cells, was determined by cycloheximide chase analysis.

HA–FAT10 was readily degraded in WT cells with a half-life of

~2 h, whereas it was stable over the 6-h chase period in rpn10∆ cells (Fig. 5a). HA–FAT10 degradation was almost completely restored in rpn10∆ cells on ectopic expression of either scRpn10 or hRpn10 (Fig. 5b). Remarkably, expression of only the VWA domain was suf- ficient to restore degradation of HA–FAT10 in cells lacking Rpn10 (Fig. 5c). Likewise, hRpn10 with mutated UIMs also restored HA–FAT10 degradation to WT-like levels.

NUB1L relies on hRpn10 for promoting the degradation of FAT10.

Because the degradation of FAT10 is accelerated by NUB1L³², and because both FAT10 and NUB1L interact with hRpn10, we investi- gated whether the stimulating effect of NUB1L on the degradation of FAT10 would depend on hRpn10 in vivo. Cycloheximide chase exper- iments were performed as described above, except that HA–NUB1L was also expressed in yeast cells. The HA–FAT10 protein remained stable in rpn10∆ yeast cells even in the presence of HA–NUB1L, but its degradation was accelerated in wild-type cells (Fig. 5d).

Following reconstitution of HA–NUB1L expressing rpn10∆ cells with either FLAG-scRpn10 or FLAG-hRpn10, degradation rates similar to that of the WT strain expressing HA–NUB1L were observed (Fig. 5e–g; Supplementary Fig. S2). The quantitative analy- sis revealed that the half-life of HA–FAT10 was reduced to less than 1 h (Fig. 5g). Collectively, these results show that the accelerating effect of NUB1L on FAT10 degradation is dependent on hRpn10.

625 WT + empty vector rpn10∆ + empty vector rpn10∆ + FLAG-scRpn10 rpn10∆ + FLAG-hRpn10 rpn10∆ + FLAG-VWA rpn10∆ + FLAG-hRpn10

(UIMs mutated)

SD-Leu 3 µg ml^–1 Canavanine 125 25 5 1X 625 125 25 5 1X

Figure 3 | Reconstitution of canavanine-resistant growth of scRpn10

yeast with hRpn10 or its VWA domain. Canavanine sensitivity plate growth assay with WT and rpn10∆ S. cerevisiae transformants. WT or rpn10∆ cells transformed with high-copy expression plasmids for FLAG- scRpn10, FLAG-hRpn10, FLAG-VWA or FLAG-hRpn10-uIms mutated or with the empty plasmid were spotted on plates with synthetic medium lacking leucine (sD-Leu) or on plates lacking leucine and arginine that contain 3 µg ml ^{− 1} canavanine. The growth assay has been repeated three times with similar results.

To confirm the earlier finding that the degradation of FAT10 is independent of ubiquitin^21,22, we analysed the degradation of lysine-deficient FAT10 in yeast and found it to be degraded as rap- idly as WT FAT10 (Fig. 5h). We noted that FLAG-hRpn10 or FLAG- VWA was also degraded over time (Fig. 5b,c) but this degradation occurred independently of HA–FAT10 (Fig. 5i).

Degradation of FAT10 via hRpn10-VWA is proteasome-depend- ent. To determine whether FAT10 is degraded by the proteasome, we investigated whether FAT10 could be stabilized by MG132 (ref. 37).

The pdr5 gene was disrupted in WT and rpn10∆ cells, as WT yeast cells readily export MG132 via the multi-drug transporter Pdr5 ren- dering them resistant to MG132. Cycloheximide chase experiments in the presence of MG132 were performed in pdr5∆ (NRY51) and pdr5∆rpn10∆ (NRY53) cells transformed with an expression plas- mid for HA–FAT10 alone or in combination with an expression plasmid for FLAG-hRpn10 or FLAG-VWA. Strikingly, addition of MG132 led to a strong stabilization of FAT10 in all the transformants tested, even in the pdr5∆rpn10∆ strain reconstituted with only the VWA domain (Fig. 6a–c). Additionally, the degradation of FAT10 was analysed in the proteasome mutant yeast strain pre1-1 (ref. 38).

Cycloheximide chase experiments were performed after transform- ing HA–FAT10 into pre1-1 and its parental strain. We observed the stabilization of FAT10 in pre1-1 strain as compared with the WT strain under non-permissive conditions (Fig. 6d). Moreover, mutat- ing the Asp11 residue in the VWA domain of human or yeast Rpn10

to Ala, which affects ubiquitin-mediated proteolysis by the 26S pro- teasome, also interferes with the degradation of FAT10 (Supplemen- tary Fig. S3). Taken together, these data strongly suggests that the degradation of FAT10 is proteasome-dependent.

FAT10 competes with NUB1L for binding to hRpn10-VWA.

As mentioned above, the N-terminal domain of FAT10 docks to NUB1L³⁴, whereas the C-terminal domain of FAT10 docks to hRpn10 (Fig. 2e). To investigate whether a trimeric complex is formed, GST–NUB1L∆UBL bound to glutathione-sepharose (reduced) beads was incubated with His-hRpn10 and FAT10 either in equimolar amount or with increasing concentrations. An increase in the binding of His-hRpn10 on the beads with the increasing con- centrations of FAT10 was clearly observed (Fig. 7a), which con- firmed that FAT10 can bind to NUB1L and hRpn10 simultaneously using distinct domains.

As both FAT10 and NUB1L can bind to the VWA domain of hRpn10 (Fig. 2f), we determined whether they can compete with each other for hRpn10 binding. To this aim, we utilized a dele- tion mutant of NUB1L lacking UBA domains, which cannot bind to FAT10 but still binds to hRpn10. Recombinant His-hRpn10 was bound to Ni-beads and recombinant GST–NUB1L∆UBA was incu- bated with these beads along with FAT10 at increasing concentra- tions. Interestingly, increasing levels of FAT10 molecules displaced NUB1L from hRpn10 (Fig. 7b), which suggests that NUB1L and FAT10 bound to the same motif on the VWA domain implying that,

Wild type

rpn10∆

28

a

b

c

17 38

2 1

28

55

26

55

28 17 38 28

55

26

55

3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

7 8 9 10111213 Fraction number

141516171819 20

7 8 9 10 11 1213 Fraction number

141516171819 20 WB: Anti-HA

WB: Anti-Rpn10

WB: Anti-pre6

WB: Anti-Rpt1

45

WB: Anti-AP I

WB: Anti-HA

WB: Anti-Rpn10

WB: Anti-pre6

WB: Anti-Rpt1 WB: Anti-AP I

40 35 30 25 20 15 10 5 0

Peptidase activity

45 40 35 30 25 20 15 10 5 0

Peptidase activity

Figure 4 | Glycerol gradient fraction analysis showing the interaction of FAT10 with the yeast proteasome. (a) Whole-cell lysate from the wild-type yeast transformed with a low-copy plasmid expressing HA–FAT10 was loaded onto a 10–40% glycerol gradient and subjected to ultra-centrifugation (see methods). The fractions were analysed by western blotting with anti-HA, anti-scRpn10, anti-aminopeptidase I (AP I), anti-pre6 and anti-Rpt1 antibodies.

HA–FAT10 migrated to higher fractions of about 600 kDa (fraction 11–12 and 15–20). AP I, which is a decamer of 600 kDa, served as size marker.

(b) Glycerol gradient fraction analysis of a whole-cell lysate from rpn10∆ cells transformed with a low-copy plasmid expressing HA–FAT10. Analysis was performed as in (a). (c) Peptidase activity analysis. The upper graph shows peptidase activity in cell lysates of wild type, the lower graph in cell lysates of rpn10∆ cells. The fractions of glycerol gradients were analysed for proteolytic activity with the fluorogenic substrate suc-LLVY-AmC (filled squares).

Treatment with mG132 reduced the peptidase activity to basal level (filled circles), which confirmed that this activity relied on the proteasome. After addition of 0.02% sDs, the peptidase activity of 20s proteasome-containing fractions was increased (filled triangles). The unspecific background activity by the substrate was subtracted.

with an excess of FAT10, NUB1L is unable to bind to hRpn10 but can still bind to Rpn1.

hRpn10 knockdown in human cells causes FAT10 accumulation.

Contrary to yeast, mammalian cells deficient in hRpn10 are not viable over extended periods of time^39,40. Therefore, the function of hRpn10 in the degradation of FAT10 and its conjugates in mam- malian cells was addressed by transiently knocking down hRpn10 with short interfering RNA (siRNA) in HEK293 cells that lead to extensive depletion of hRpn10 (Fig. 8a) but did not affect their via- bility during the course of the experiment. The steady-state level of monomeric FAT10, as well as FAT10 conjugates, was significantly increased in cells treated with hRpn10 siRNA as compared with

cells treated with control siRNA (Fig. 8a,b). In HEK293 cells stimu- lated with IFN-γ and TNF-α to induce endogenous FAT10 expres- sion, there was no accumulation of FAT10 or FAT10 conjugates when normalized to β-actin expression (Fig. 8c). However, real- time RT–PCR analysis revealed that treatment of cytokine-induced cells with hRpn10 siRNAs consistently reduced endogenous FAT10 messenger RNA levels by ~60% when compared with cells treated with control siRNA (Fig. 8d). When we normalized the levels of monomeric or substrate-conjugated endogenous FAT10 protein on the level of FAT10 mRNA, we confirmed an accumulation of FAT10 and its conjugates after knockdown of hRpn10 (Fig. 8d).

Recently, we have identified the ubiquitin-activating enzyme UBE1 as a substrate of FAT10ylation, and we tested whether the

b c

d

a ^WT

+ Empty vector 0

17 55 43

0.5 1 2 4 6 0 0.5 1 2 4 6

0 0.5

+ FLAG-scRpn10 + FLAG-hRpn10 1 2 4 6 0 0.5 1 2 4 6 17

(h) FAT10 Pgk1

(h) FAT10 Pgk1 Rpn1055

(h) FAT10

120

Pgk1 Rpn1055 rpn10∆

WT

+ HA–NUB1L

Nub1L

Nub1L 0

17 55 43 70

17 55 43

70 28 38 0.5 1 2 4 6 0 0.5 1 2 4 6 (h)

FAT10 Pgk1 rpn10∆

WT

+ HA-FAT10 + HA-FAT10 K0 0

17 55 43

55

43

55 43 26

17

17 0.5 1 2 4 6 0 0.5 1 2 4 6

0 0.5 1 2 4 6 0 0.5 1 2 4 6 (h)

FAT10 Pgk1

h 0 0.5 1 2 4 6 0 0.5 1 2 4 6 h

hRpn10 Pgk1

Pgk1 FAT10

FAT10

rpn10∆ rpn10∆

0 0.5 + HA-NUB1L+

FLAG-scRpn10

+ HA-NUB1L+

FLAG-hRpn10

+ FLAG-hRpn10 + FLAG-hRpn10 + HA-FAT10 1 2 4 6 0 0.5 1 2 4 6

rpn10∆ rpn10∆

rpn10∆ rpn10∆

+FLAG-VWA

VWA +FLAG-VWA +

HA-FAT10

rpn10∆ rpn10∆

h i

e f g

55 43 38 28

0 0.5

+ FLAG-VWA + FLAG-hRpn10 (UIMs mutated) 1 2 4 6 0 0.5 1 2 4 6 17

(h) FAT10 Pgk1 VWA/

hRpn10 rpn10∆ rpn10∆

5543 38

100 80 60 40 20 0

120

100 75 50 25

0

0 0.5 2 6

Chase time (h)

0

100 80 60 40 20 0

0.5 2 6

Without FAT10 With FAT10

Chase time (h)

0 0.5 2 6

Chase time (h)

0 0.5 2 6

Chase time (h)

Percentage of FAT10 remaining Percentage of FAT10 remaining

Percentage of hRpn10 remaining

100 75 50 25 0 Percentage of VWA remaining FAT10

FAT10 K0

0 0.5 2 6

Chase time (h) 100

75 50 25 Percentage of remaining protein 0

Without FAT10 With FAT10

Figure 5 | Reconstitution of FAT10 degradation in Rpn10-deficient yeast by the VWA domain of hRpn10. Cycloheximide chase analysis of FAT10 stability in WT and rpn10∆ cells reconstituted with hRpn10 or variants and/or nuB1L. HA-tagged FAT10 and nuB1L were visualized by anti-HA western blotting whereas FLAG-tagged proteins, scRpn10, hRpn10, VWA and hRpn10-mutated uIms were detected via FLAG-specific antibodies. Pgk1 served as a loading control. (a) HA–FAT10 degradation in WT and rpn10∆ cells. The experiment was repeated three times with similar outcome (quantitation is shown in (f)). (b) FAT10 degradation in rpn10∆ yeast reconstituted with FLAG-scRpn10 (left) or FLAG-hRpn10 (right). (c) FAT10 degradation in rpn10∆

yeast reconstituted with FLAG-VWA (left) or FLAG-hRpn10-uIms mutated (right). (d) Accelerated FAT10 degradation in the presence of nuB1L in wild-type but not rpn10∆ yeast. (e) Accelerated FAT10 degradation in the presence of nuB1L in rpn10∆ yeast reconstituted with FLAG-scRpn10 (left) or FLAG-hRpn10 (right). (f, g) Quantitation of HA–FAT10 cycloheximide chase experiments in the absence of nuB1L (f) and in the presence of nuB1L (g); WT + empty vector (white), rpn10∆ + empty vector (light grey), rpn10∆ + scRpn10 (hatched), rpn10∆ + hRpn10 (dark grey), rpn10∆ + VWA (dotted), rpn10∆ + hRpn10 with uIms mutated (black). Values represent the mean ± s.e.m. from three experiments (f) and the means ± the range from two (g) experiments (h) Degradation of FAT10 in yeast is independent of ubiquitin. Cyloheximide chase was performed after transformation of wild-type yeast with either FAT10 or lysine-deficient FAT10, where all lysines are replaced by arginines (FAT10 K0). The graph represents the mean value ± range from two sets of experiments. (i) no effect of FAT10 on the degradation of full-length hRpn10 (top) or the isolated VWA domain (bottom) of hRpn10.

Cycloheximide chase experiments were performed for 6 h in rpn10∆ yeast reconstituted with FLAG-hRpn10 or FLAG-VWA with and without HA–FAT10 as indicated. Pgk1 western blots served as loading controls. The anti-FLAG western blots and their quantitative evaluations show that the degradation of hRpn10 and VWA is independent of FAT10 expression.

UBE1–FAT10 conjugate would accumulate in HEK293 cells upon hRpn10 knockdown. Depletion of hRpn10 led to a twofold increase in the amount of UBE1–FAT10 conjugate (Fig. 8e, f). These results suggest that hRpn10 functions as a receptor for FAT10 and its conjugates and targets them for proteasomal degradation also in mammalian cells.

Discussion

Degradation of proteins by the proteasome is regulated at several levels. One level is the docking step to the 26S proteasome, which in the ubiquitin system, is mediated by binding of poly-ubiquitin chains to the proteasome subunits Rpn10 or Rpn13 or to soluble ubiquitin receptors, which in turn bind to Rpn1 or Rpn2 (ref. 1).

Interestingly, these two binding modes seem to mutually affect each other, as the degradation of poly-ubiquitylated Sic1 at the isolated 26S proteasome via Rad23 was shown to be facilitated by binding of

the VWA domain of the Rpn10 subunit to the proteasome¹⁹. Similar evidence regulating the pace of proteasomal degradation has been observed for the FAT10 system where NUB1L accelerated FAT10 degradation about eightfold³². This stimulating effect relied on the N-terminal UBL domain of NUB1L required for proteasome bind- ing and was independent of the three UBA domains of NUB1L, which bind to FAT10 (ref. 34). FAT10, therefore, resembles ubiqui- tin in that it can bind directly to the proteasome or can become teth- ered to the proteasome via the UBL–UBA protein NUB1L. To better understand how NUB1L influences the rate of FAT10 degradation, we set out to identify the subunit(s) of the RP which bind to FAT10.

Rpn10 was the only RP subunit found to interact with FAT10 in our yeast two-hybrid analysis.

The finding that FAT10 and NUB1L bind to the same subunit raised the question how NUB1L may act as a facilitator of FAT10- mediated degradation. One possibility was that FAT10 and NUB1L

pdr5∆

– MG132 + CHX + MG132 + CHX 0 0.5 1 2 4 6

FAT10 Pgk1 0 0.5 1 2 4 6 (h) 17

55 43

a

pdr5∆rpn10∆ + FLAG-hRpn10 – MG132 + CHX + MG132 + CHX 0 0.5 1 2 4 6

FAT10 Pgk1 hRpn10 0 0.5 1 2 4 6 (h) 17

55

55 43

b

pdr5∆rpn10∆ + FLAG-VWA – MG132 + CHX + MG132 + CHX 0 0.5 1 2 4 6

FAT10 Pgk1 VWA 0 0.5 1 2 4 6 (h) 17

55 43 26

c

0 0.5 1 2 4 6

FAT10 Pgk1

d pre1–1 (38 °C)

0 0.5 1 2 4 6 (h) 17

WT (38 °C)

43 55

pdr5∆ (– MG132) pdr5∆ (+ MG132) 100

80 60 40 20

0 0

Percentage of FAT10 remaining

0.5 2 6

Chase time (h)

pdr5�rpn10∆ + hRpn10 (– MG132) pdr5�rpn10∆ + hRpn10 (+ MG132)

Percentage of FAT10 remaining 100 120 80 60 40 20 0

0 0.5 2 6

Chase time (h)

pdr5�rpn10∆ + VWA (+ MG132) pdr5�rpn10∆ + VWA (– MG132)

Percentage of FAT10 remaining 100 120 140

80 60 40 20

0 0 0.5 2 6

Chase time (h) WTpre1–1

Percentage of FAT10 remaining 100 120 80 60 40 20

0 0 1 4 6

Chase time (h)

Figure 6 | The degradation of FAT10 in vivo is dependent on proteasome activity. (a) HA–FAT10 cycloheximide chase analysis in pdr5∆ yeast cells in the presence or absence of the proteasome inhibitor mG132 with quantitative analysis of the densities of HA–FAT10 bands in the western blot.

(b) Cycloheximide chase analysis with pdr5∆rpn10∆ cells ectopically expressing FLAG-hRpn10; western blots on the bottom monitor FLAG-hRpn10 levels. (c) Cycloheximide chase analysis with pdr5∆rpn10∆ cells ectopically expressing FLAG-VWA; western blots on the bottom monitor FLAG-VWA expression. (d) Cycloheximide chase analysis of the degradation of HA–FAT10 in wild type (WCG4a) and the proteasome subunit mutant pre1-1 yeast strain with quantitative analysis of HA–FAT10 western blots. Pgk1 western blots served as loading controls in all panels. The graphs represent the mean value ± range from two independent sets of experiments.

bind to different domains of Rpn10. It was therefore surprising that both of them bound to the same domain. Even more surprising was that these two ubiquitin-like proteins did not bind to the UIM domains of Rpn10, like ubiquitin, but instead to the VWA domain, which seems to be a novel UBL-binding domain (Fig. 2f). VWA domains contain a β-sheet sandwiched by multiple α-helices and they usually bind to metal ions via a metal ion-dependent adhesion site. It was hypothesized that the VWA domains mediate protein–

protein interactions involved in the assembly or function of multi- protein complexes^7,36. We have extensively tried to identify the residues within the VWA domain that are involved in FAT10- and NUB1L-binding by mutagenesis. However, several combined point mutations were unsuccessful in abolishing the binding to NUB1L or FAT10, whereas deletion variants could not be stably expressed.

The human Rpn10 subunit was able to reconstitute growth of rpn10∆ S. cerevisiae cells in canavanine sensitivity assays (Fig. 3).

Rpn10 is required to keep the base and the lid of the RP firmly asso- ciated^6,19 and our transcomplementation data with human Rpn10 suggest that the human orthologue can serve this structural func- tion also in the context of the yeast 26S proteasome. The reconsti- tution of Rpn10-deficient yeast with human Rpn10 encouraged us to test whether FAT10 degradation could be reconstituted in yeast.

The reconstitution of FAT10 degradation in yeast seemed justified as Rpn10 has been shown to be essential in mouse cells⁴⁰ and is required for normal degradation of ubiquitin-conjugates in Drosophila³⁹. We could show that FAT10 is degraded in yeast in a Rpn10-dependent manner (Fig. 5). The fact that FAT10 degradation was supported by both yeast and human Rpn10 is consistent with the finding that not the UIMs (which differ in number between human and yeast), but the VWA domain of Rpn10 is required for FAT10 degradation.

This system further allowed us to show that the VWA domain suf- ficed to robustly reconstitute FAT10 degradation (Fig. 5c), which is in accordance with our binding studies (Fig. 2f). Interestingly, the VWA domain could largely restore wild-type-like growth of Rpn10- deficient yeast on canavanine-containing plates (Fig. 3). Consist-

ently, mice bearing only a VWA domain survive longer than mice that completely lack mRpn10 (ref. 40). The knockdown of hRpn10 in human HEK293 cells led to an accumulation of FLAG–FAT10 and endogenous FAT10 and bulk FAT10 conjugates (when normal- ized to the level of FAT10 mRNA), and the UBE1–FAT10 conjugate by 30–50% (Fig. 8), suggesting that hRpn10 serves as a FAT10 recep- tor also in humans, although the effect was not as strong as found after rpn10 deletion in yeast. This difference is most likely due to residual hRpn10 protein in the siRNA-treated human cells.

Previously the only subunit known to interact with NUB1L was hRpn10 and it was shown that the C terminus of NUB1L binds to hRpn10 (ref. 33). Another study reported that the association of NUB1L with the 26S proteasome occurs via the N-terminal UBL domain³⁴. We show here that the UBL domain of NUB1L binds to hRpn1 as well as hRpn10 of the 26S proteasome in accordance with the latter report (Fig. 2d). Interestingly, we could also show that the C-terminal UBL domain of FAT10 binds to hRpn10 (Fig. 2e), which leaves the N-terminal UBL domain of FAT10 free for docking to the UBA domains of NUB1L³⁴ resulting in formation of a trimeric NUB1L–FAT10–Rpn10 complex (Fig 7a).

One may speculate about the biological consequences of FAT10 being degraded slowly in the absence of NUB1L and faster in its presence. It is possible that different FAT10ylated substrates require different rates of proteasome-mediated degradation and that only those are bound by NUB1L, which require rapid degradation. A striking difference between ubiquitin and FAT10 is that ubiqui- tin is recycled with the help of numerous ubiquitin-deconjugat- ing enzymes whereas such enzymes have not been found to date for FAT10 that seems to be degraded along with its substrates^21,22. The overexpression of FAT10 is toxic to cells²⁷ and it is possible that NUB1L contributes to keeping FAT10 at a low level.

The key findings of this study can be exploited to suggest two models for how NUB1L accelerates the degradation of FAT10.

FAT10 might directly bind to the VWA domain of hRpn10 and become degraded at a slow rate in the absence of NUB1L (Fig. 8g).

GST-NUB1L∆UBL

GST-NUB1L∆UBA GST

+ + + 1X 2X 4X+ + + + + + + + + – – – – – –

– – – – –

– – – – –

Anti-His-HRP

Anti-His-HRP IP: GSH beads

Blocked Blocked

NUB1L NUB1L

N C VWA UIMS

hRpn10

hRpn10 FAT10

FAT10

17 55

IP: Ni-beads

Anti-FAT10

Anti-FAT10

Anti-GST

Anti-GST

26 34 43 5572 95

55

a b

17 72 FAT10

FAT10 His-hRpn10

His-hRpn10 + + + + + +

+ + + + + + + +1X 2X 4X

Figure 7 | Trimeric complex formation and competition between FAT10 and NUB1L for hRpn10. (a) Trimeric complex formation. 60 pmol of recombinant GsT or GsT–nuB1L∆uBL was precipitated on GsH-sepharose beads and incubated with 30 pmol of His-hRpn10 (lane 2) either alone or together with 60 pmol or increasing molar excess of FAT10 (lane 4, 5 and 6). After incubation for 2 h at 4 °C, the beads were washed and the interaction was analysed by western blotting with anti-His-HRP, anti-FAT10 and anti-GsT antibodies. (b) 30 pmol of recombinant His-Rpn10 was precipitated on ni-beads and incubated with 60 pmol of GsT–nuB1L∆uBA in the absence or presence of FAT10 (lane 4–8). FAT10 was added in increasing molar excesses as indicated in lane 7 and 8. The binding of nuB1L∆uBA during the course of competition was analysed by western blotting. The experiments have been repeated at least twice with similar results.

If NUB1L binds to the Rpn1 subunit, it may act like a soluble

‘FAT10-receptor’, and transfer-bound FAT10 and its conjugated substrates to the hRpn10 subunit of the proteasome to be degraded faster (Fig. 8g, transfer model). Alternatively, NUB1L binding to Rpn1 could induce a conformational change in the hRpn10 subunit (Fig. 8g, facilitator model), which allows FAT10 and its conjugated substrates to be more rapidly degraded.

Methods

Yeast strains and plasmids. Yeast strains and plasmid cloning are described in Supplementary Tables S1 and S2, respectively.

GST-pull-down-assay. GST, GST–FAT10, GST–NUB1L, GST–NUB1L∆UBA, GST–NUB1L∆UBL (Supplementary Table S2) and His-hRpn10 (ref. 12) (kindly provided by Patrick Young, Stockholm University) were expressed in Escherichia coli strain B834(DE3)pLysS as described³². Glutathione sepharose 4B beads (GE Healthcare) were incubated with GST, GST–FAT10 or GST–NUB1L along with purified His-hRpn10 in incubation buffer (20 mM Tris–HCl, pH 7.4, 0.1% Tri- tonX-100) overnight at 4 °C and washed 5 times with incubation buffer. For other experiments, His-hRpn10, His-UIM1 + UIM2 domains and His-VWA domain were in vitro transcribed and translated using TNT T7 coupled reticulocyte lysate system (Promega) and purified GST, GST–FAT10 or GST–NUB1L bound on the GSH beads was added. Proteins were eluted by boiling at 95 °C for 5 min with SDS buffer, resolved by SDS–PAGE and subjected to western blot analysis using anti-His antibody (1 µg ml ^{− 1}, Sigma) or autoradiography.

Control siRNA

a

e g

f

b c d

hRpn10 siRNA Control siRNA

FAT10 conjugates hRpn10 siRNA

Relative FAT10 expression normalized to GAPDHMonomeric FAT10 normalized to FAT10 mRNA FAT10 conjugates normalized to FAT10 mRNA

FLAG (FAT10) hRpn10

β-Actin

*

*

IP: FLAG

FLAG -FAT10 monomer (load)

FLAG -FAT10 conjugates (IP) WB: FLAG

+ +

FAT10 hRpn10 FAT10 β-Actin

–

– +

kDa

100

175 200

150 100 50 0 150

125 100 75 50 25 0 75 50 25 0 – 75

– 50 – 37 – 25– 20

Load

– –

+ kDa – 100

200180 160140

FLAG-FAT10 (load) normalized to β-actinFLAG-FAT10 (IP) normalized to β-actinUBE1-FAT10 normalized to β-actin 120100

8060 4020 0

140

IP: FAT10 WB: FAT10

*

120 100 80 60 40 20 0

300 250 200 150 100 50 0 – 75

– 50 – 37 – 25 – 20

Control siRNA hRpn10siRNA

Control siRNA

ControlsiRNA hRpn10

siRNA

hRpn10siRNA

ControlsiRNA hRpn10siRNA

Control

siRNA hRpn10siRNA ControlsiRNA hRpn10siRNA Load

hRpn10

Fast degradatio n

S FAT10

Fast degradation Facilitato r Transf

er of FAT10

FAT10 N

UBL UBL

UBA1 UBA2

UBA3

UBA1 UBA2

C N UBA3

C S

hRpn10

20s

20s 20s

hRpn10

hRpn10 hRpn10 Rpn1

Rpn1

Rpn1 UIM2

N C

FAT10 S

UIM1

UIM1

DegradatioSlow n UIM2 keaW

Strong VWA

VWA

UBL UBA1 UBA2 UBA3 NUB1L Control siRNA

HA-UBE1- FLAG-FAT10

hRpn10 FLAG (FAT10)

HA (UBE1)

Load

β-Actin

*

IP: HA WB: FLAG

hRpn10 siRNA + –

– +

kDa –150 –100 –75

–50

*

Figure 8 | Knockdown of hRpn10 in HEK293 cells causes accumulation of FAT10 and FAT10 conjugates. (a) HEK293 cells were transfected with hRpn10 or control siRnA and His-3×FLAG-FAT10 (FLAG-FAT10). Lysates were immunoprecipitated with anti-FLAG-sepharose. A western blot of immunoprecipitated FLAG-FAT10 and indicated proteins in lysates (load) are shown. Asterisks mark unspecific signals. (b) Quantification of monomeric and conjugated FAT10 in (a) normalized to β-actin. significance was calculated with paired two-tail P-value test. Values represent the mean ± s.e.m. of 9 (monomeric FAT10) and 7 (FAT10 conjugates) experiments. (c) Accumulation of endogenous FAT10 upon downregulation of hRpn10. HEK293 cells were treated with control or hRpn10 siRnA and 24 h later with IFn-γ/TnF-α for extra 24 h. FAT10 immunoprecipitates were analysed by western blot with a polyclonal FAT10-reactive antibody (top) and expression levels of hRpn10, endogenous FAT10, and β-actin in lysates are shown (below). one of three experiments with similar outcome is shown. (d) Induction of endogenous FAT10 expression was measured by real-time RT–PCR, RnA levels were normalized to GAPDH. The mean of three independent experiments is shown. signals for monomeric (load) and conjugated FAT10 (IP) were normalized to FAT10 mRnA levels. (e) Western blot showing accumulation of uBE1–FAT10 conjugate upon downregulation of hRpn10. HEK293 cells were transfected with hRpn10 or control siRnAs and His-3×FLAG-FAT10 (FLAG-FAT10) and HA-uBE1 plasmids. Western blots of HA immunoprecipitates (top) and indicated proteins in total lysates (load) are shown. A representative experiment out of three is shown. (f) signals of three experiments as in (e) were quantified and normalized to β-actin. Values represent mean ± s.e.m. of three experiments. (g) models of FAT10 and nuB1L-mediated degradation by the 26s proteasome. (Top) slow degradation of FAT10 conjugates by the 26s proteasome when interacting with the VWA domain of hRpn10 in the absence of nuB1L. (Bottom left) Transfer model showing accelerated degradation of FAT10 when nuB1L interacts with hRpn1 and the n-terminal uBL domain of FAT10, which is subsequently transferred to hRpn10. (Bottom right) Facilitator model: FAT10 docks onto hRpn10 while its degradation is facilitated by binding of nuB1L to hRpn1 leading to conformational changes in both subunits.